Modelling persistence in the conditional mean of inflation using the ARFIMA process with GARCH and GJRGARCH innovations: The case of Ghana and South Africa

This paper contributes to the debate on inflation persistence by extending an ARFIMA process with GARCH and GJR-GARCH models to describe the time-dependent heteroskedasticity and persistence in the conditional mean of Consumer Price Index (CPI) inflation series of Ghana and South Africa, under three distributional assumptions (i.e., Normal, Student-t and Generalised Error Distributions). ARFIMA-GJR-GARCH under Generalised Error Distribution and Student-t Distribution respectively, provided the best fit for modelling the time-dependent heteroskedasticity and persistence in the conditional mean of CPI inflation rate of Ghana and South Africa. The results from the study provided evidence of persistence, mean reverting though, and asymmetric effect of economic shocks on the conditional mean of CPI inflation rate of the two countries. These results would, therefore, be useful to both countries in making good portfolio decisions, assessing the efficacy of a monetary policy or programme meant to control inflation persistence and also serving as a tool for detecting volatility and its impact, for the Ghanaian and South African inflation rates and their economies at large.Keywords: CPI Inflation; Fractional integration; Persistence; Conditional mean; ARFIMA; GARCH; GJR-GARCH model